Works with the agents you already use

One command per client. Your existing config is left alone; OpenLLM adds fallback chains, MCP code search and memory, and one usage dashboard.

Start free See pricing

Works with Claude Code

Run Claude Code through one gateway

Keep Claude Code exactly as it is. Add fallback chains for rate limits, models from your other providers, MCP code search and a usage dashboard with one command.

Works with Codex

Run Codex through one gateway

Point the Codex CLI at OpenLLM for fallback chains, a shared model catalog across devices, MCP code search and per-model cost reporting.

Works with DeepSeek Harness

Use DeepSeek Harness with OpenLLM

Add the OpenLLM bundle to a DeepSeek Harness profile and route it through your local gateway, fallback chains included.

Works with ChatGPT

Start the ChatGPT Mac app wired to OpenLLM

Launch the ChatGPT desktop app with OpenLLM behind it, so its requests show up in the same usage dashboard as every other client.

Works with Grok Build

Run Grok Build through one gateway

Use xAI's coding agent with OpenLLM fallback chains, MCP code search and one usage view across every client you run.

Works with Hermes

Run Hermes Agent on OpenLLM

Give Nous Research's autonomous agent every model in your catalog, fallback chains and cost reporting through one OpenAI-compatible endpoint.

Works with OpenCode

Run OpenCode with every model you have

OpenCode on one endpoint: your API-key providers, fallback chains, MCP code search and memory, and a dashboard of what each model costs.

Works with Raycast

Bring your models to Raycast AI

Make the models in your OpenLLM catalog available inside Raycast, with fallback chains and usage tracking behind them.